-
Notifications
You must be signed in to change notification settings - Fork 1.9k
Description
Is your feature request related to a problem or challenge?
DDL extensions
Describe the solution you'd like
Currently, CREATE/DROP schema and table operations in DataFusion are in-memory only(The tutorial said). The CatalogProvider::register_schema and SchemaProvider::register_table APIs allow extending the behavior of these operations—but they are synchronous and they can only build MemTable.
In production scenarios, metadata is typically managed by a metastore or external catalog service. It would be great to support creating and dropping schemas, tables (and even catalogs) via such services. Ideally, these operations should be asynchronous and expose elements that are necessary for building customized TableProvider, SchemaProvider and CatalogProvider.
I’m curious if this has been discussed before—could the community please share any relevant context? Thanks!
Describe alternatives you've considered
Just use SchemaProvider::register_table to integrate catalog services. However, I think the solution is somehow limited becase the backend table provider is based on memory and can never be refreshed
Additional context
No response